As an alternative to the Gordon Growth models, this section describes a quantitative valuation model based on the value driver formula presented by McKinsey (2015, p. 31). The advantage of this model is that it illustrates the relationship between growth, free cash flow and returns on invested capital. To grow, a company must invest - especially if it is a capital-intensive business.
This relationship exposes how slowing growth assumptions can have a significant positive impact on free cash flow because less cash must be reinvested in the business.
𝐶𝑜𝑛𝑡𝑖𝑛𝑢𝑖𝑛𝑔 𝑉𝑎𝑙𝑢𝑒𝑡= 𝑁𝑂𝑃𝐴𝑇𝑡+1× (1 − 𝑔 𝑅𝑂𝑁𝐼𝐶) 𝑊𝐴𝐶𝐶 − 𝑔
NOPAT: Net operating profit after taxes. McKinsey (2015, p. 31) apply NOPLAT (net operating profit less adjusted taxes). Later, we will explain why.
g: growth of NOPAT in all future periods.
RONIC: Return on new invested capital, or rather, the return on the next dollar invested.
g/RONIC: Defines the investment rate (IR) or the portion of NOPAT reinvested in the business.
With Gordon's Growth formula, analysts are prone to ignore the link between growth and free cash flow. If the growth assumption in steady state is lower than the growth at the end of the explicit forecast period (which it often is), it should not be necessary for the firm to keep reinvesting the same portion of its operating income, and this would make the free cash flow higher. If steady-state FCFF is based on the most recent fiscal year (as we do in one of our Gordon Growth models) or is based on the cash flow in the final year of an analyst’s explicit forecast period, the FCFF in perpetuity can often be too low and underestimate the terminal value significantly (McKinsey, 2015, p. 250).
When assuming a lower growth in the value driver formula, the investment rate (g/RONIC) will decrease while the cash flow (NOPAT × (1 - g/RONIC) will increase. For this reason, the value driver formula can prove superior when companies have not yet reached steady state.
Morningstar applies a similar formula when calculating terminal value but always assumes RONIC to equal cost of capital (WACC). This implies that all new investments and growth will not generate any value, which deletes growth and the investment rate from the formula (see Section 2.3).
Page 37 NOPAT and tax
Net operating profit after tax (NOPAT) is computed as follows:
𝑁𝑂𝑃𝐴𝑇 = 𝐸𝐵𝐼𝑇 × (1 − 𝑡𝑎𝑥 𝑟𝑎𝑡𝑒)
McKinsey’s NOPLAT also adds the change in deferred taxes and excludes amortizations by using EBITA instead of EBIT. To avoid large year-to-year fluctuations in deferred taxes, that might affect our steady-state estimates, we apply NOPAT. Using EBIT instead of EBITA means that amortizations of acquired intangibles are also subtracted from operating profit, but McKinsey (2015, p. 393) argues that such noncash expenses should be excluded. The argument is that only acquired intangibles are capitalized and amortized, whereas internally developed intangible assets, such as brands, are expensed as SG&A when they are created (due to accounting standards). Thus, if a company acquires an intangible asset and then replenishes the asset through internal investments, its EBIT will be penalized twice; once through selling, general, and administrative expenses (SG&A) and again through amortization (McKinsey, 2015, p. 393).
Yet, we would like to point out that this argument goes both ways, and EBITA can also overstate operating profit. Excluding amortizations would favor companies where acquiring brands, patents or client relations is a consistent part of the business as an alternative to developing these assets internally. Big pharma is an example where mergers and acquisitions are a consistent part of operations, and where acquired patents become recognized on the balance sheet. In such cases, EBITA will not be penalized by amortizations, and at the same time, SG&A will appear lower because intangibles are purchased externally instead of being developed internally and expensed (Nissim, 2017a, p. 17). In this case, EBITA would overstate operating profit.
When estimating both free cash flows and operating profit, investments are necessary not only to maintain existing assets but also to grow. In this context, depreciations can be seen as a proxy for the cost of maintaining the existing asset base (Nissim, 2017a, p. 12). In the same way, amortizations approximate the cost of maintaining intangible assets (if the value of brands and patents erode over time). Since companies can determine whether to replace their assets this year or the next, excluding depreciations and amortizations could be optimal in the near term. But for estimating long-term profitability and free cash flows - which we strive to do in our terminal value models - D&A should not be ignored (Nissim, 2017a, p. 12). In the end, whether EBITDA, EBITA or EBIT best describe operating profit depends on the individual firm, but since we do not have the luxury of making such individual distinctions in a quantitative model, we lean towards EBIT for the reasons listed above.
Page 38 It is challenging to estimate the tax rate on operating income. A common approach, which we follow, is to apply the effective tax rate (the ratio of the income tax expense to earnings before tax) as follows:
𝑇𝑎𝑥 𝑟𝑎𝑡𝑒 = 𝐼𝑛𝑐𝑜𝑚𝑒 𝑇𝑎𝑥𝑒𝑠
𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑏𝑒𝑓𝑜𝑟𝑒 𝑡𝑎𝑥 (𝐸𝐵𝑇)
The effective tax rate represents the weighted average tax rate on operating income and other items included in the earnings before tax such as interest income/expenses. If the tax on these items is different than on the operating profit, the effective tax rate can be a poor proxy (Nissim, 2017, p. 39). Transitory items such as unrecognized tax benefits, unreserved prior periods tax payments and changes in tax rates contribute to the volatile nature of the effective tax rate (Nissim, 2017, p. 40).
An alternative would be to apply the marginal federal and state tax rate in the U.S. such as Kaplan & Ruback (1994, p. 9) in each given year, but the S&P 500 constituents are mostly international firms with operations all over the world that typically pay lower tax rates on foreign earnings (McKinsey, 2015, p. 380). Using the same tax rate across firms would also eliminate the fundamental differences that allow some industries to benefit from R&D tax credits or the ability to optimize their tax efficiency by placing operations and assets in low-tax countries.
In our model, we apply the effective tax rate on EBT for simplicity. If the tax rate is larger than 50% or lower than 10%, which we interpret as unsustainable in the long term, we ignore this value and instead apply a median historical tax rate of up to 5 prior years.
We apply the most recent fiscal year’s NOPAT but also test historical averages of NOPAT as an input in the value driver formula. A historical average would have a negative effect on the valuation of growing businesses and high-growth sectors such as technology, since their historical operating profits are typically lower than their current.
Since we apply a raw, unadjusted EBIT we will not risk overstating steady-state NOPAT by excluding “one-time” losses such as impairments and restructuring costs, which are not commonly forecasted by qualitative analysts - although these items can sometimes be recurring in nature (Nissim, 2017, p. 27). Special items are most often negative (Johnson, Lopez &
Sanchez, 2011), so excluding them will often result in higher operating profits. Since we do not exclude special items, we run the risk of understating the operating profit of the continuing business - resulting in lower valuations. In the case of large positive special items, we may risk overestimating steady-state NOPAT by not excluding them.
Page 39 Invested Capital
To compute returns on new invested capital (RONIC), we must determine a quantitative proxy for invested capital with the data points available in Morningstar Direct. As we have already estimated NIBD in Section 3.1, we can compute invested capital as the net financing (or capital) on the right-hand side of the balance sheet (Sørensen, 2012, p. 159):
𝐼𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 = 𝑁𝑒𝑡 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝐷𝑒𝑏𝑡 + 𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦
Total equity includes minority interests. One important item to outline is goodwill, which is implicitly included in our proxy of invested capital. This is consistent with also including goodwill impairments in NOPAT. The formula above should be identical to the net assets approach on the left-hand side of the balance sheet, where operating liabilities are subtracted from operating assets (Callahan & Mauboussin, 2014). The net assets approach is preferable in practice but harder to determine quantitatively with the data points available in Morningstar Direct.
Return on new invested capital (RONIC) RONIC can be measured as:
𝑅𝑂𝑁𝐼𝐶 = 𝑁𝑂𝑃𝐴𝑇𝑡− 𝑁𝑂𝑃𝐴𝑇𝑡−1
𝐼𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑡− 𝐼𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑡−1
RONIC determines the marginal operating profit after tax as a percentage of the marginal capital invested in a given period. RONIC measures the returns generated when a company invests its capital to create new value from core operations.
The expected future RONIC should be consistent with the firm’s competitive environment (McKinsey, 2015, p. 250). Economic theory suggests that competition will erode and eliminate any abnormal returns over time. This would be an argument for assuming steady state RONIC to equal WACC in competitive industries. If the return on new investments is equal to the capital costs, then growth will not be value accretive but instead value neutral with a net present value of zero. Higher growth will increase NOPAT but is offset by larger investments. Over time, this will dilute the return on invested capital (ROIC) until it reaches WACC (McKinsey, 2015, p.
262). Making such an assumption in the long term can be reasonable but might be overly conservative in a single-period valuation model - especially for firms with sustainable competitive advantages such as network effects, brands, or patents not captured on the balance sheet as argued by McKinsey. Similarly, asset-light operating models can demand long-term sustainable and high returns on invested capital above the level of WACC, because the amount of invested capital on their balance sheet is lower.
Page 40 Due to the marginal changes from year to year, the nature of RONIC have been very volatile in our dataset over time. For this reason, applying a historical RONIC in the terminal value formula is not ideal for stable valuations. RONIC is greatly influenced by noise when large investments take place, and the increased profits as a direct consequence of new investments may take years to materialize. The immediate effect of new investments on operating profit can also be severe due to straight line depreciations. Qualitative assessments by equity analysts such as RONIC converging (fading) towards WACC over time or assuming a level between a firm’s current ROIC and WACC in a competitive advantage period (CAP) can prove valuable depending on the individual firm in question (Mauboussin & Johnson, 1997). But for a quantitative approach, we will instead look towards ROIC, which is a more stable measure of profitability, or we will simply assume RONIC to equal WACC in one of the variations of our value driver models.
Return on invested capital (ROIC)
Return on invested capital (ROIC) can differ materially from RONIC due to the law of diminishing returns, competition and attractive investment opportunities becoming scarcer over time. Starbucks’ flagship store in Seattle might for example drive attractive returns (ROIC) for many years to come, but as it becomes more and more difficult to find new attractive locations, the return on new stores over time (RONIC) will decline. Still, due to the volatile nature of RONIC and for simplicity, we will apply a less conservative assumption; that a firm’s RONIC becomes equal to its historical ROIC in the value driver formula.
ROIC is a key driver of terminal value, and recent research has indicated that median ROIC over a business cycle dominates the firm’s current ROIC at predicting steady-state profitability (Nissim, 2017, p. 22). Companies earning high returns tend to experience gradually falling ROIC over the succeeding 15 years, while companies earning low returns tend to see them rise over time - thus exhibiting some mean reversion (McKinsey, 2015, p. 110 and Nissim, 2017, p. 17).
According to the studies mentioned, the best performing companies can generally sustain their abnormal returns for more than 10 years. Thus, assuming ROIC to approach WACC in valuations is overly conservative for the typical company with sustainable competitive advantages.
The ROIC of individual companies also trends toward their industry medians over time but is relatively persistent (McKinsey, 2015, p. 105). A historical median of industry ROIC tends to be superior when forecasting steady state profitability in comparison to a firm’s current ROIC, historical median ROIC and the industry’s current ROIC (Nissim, 2017, p. 22). However, this result, is most evident in the longer term with more than three years of explicit forecasts (with terminal value estimated as of T = 3). Since the valuation models in this thesis calculates an immediate terminal value with no explicit forecasts, these results might indicate that applying median industry ROIC is less advantageous. We’ve decided to keep it simple and not apply industry ROIC in our quantitative valuation models.
Page 41 However, since we utilize 10 different sector-WACC in our valuations, we will be setting RONIC equal to the sector WACC in one of our value driver models. We estimate ROIC as follows:
𝑅𝑂𝐼𝐶 = 𝑁𝑂𝑃𝐴𝑇
(𝐼𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑡+ 𝐼𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑡−1)/2
By using the average invested capital at the beginning and end of the year as stated in the formula, we arrive at a more precise estimate of the return on invested capital across a whole fiscal year (Plenborg, Petersen & Kinserdal, 2017, p. 142-148).
The estimated ROIC on the constituents of the S&P 500 is mostly not a volatile number from one year to another and outliers are rare. The estimate is quite stable but with a downward trend since 1993 which corresponds with the erosion of competitive advantages and abnormally attractive investment projects over time (McKinsey, 2015, p. 250). This is illustrated in Figure 3.1 below. Due to this trend, using a historical median of a company’s ROIC will tend to yield a higher fair value compared to simply applying ROIC from the last fiscal year in our data sample.
Since ROIC also exhibits some clear signs of cyclicality (consistent with prior research by Nissim, 2017, p. 17) during the Dot-com bubble and financial crisis, these observations can offset this effect.
Figure 3.1: Returns on invested capital (ROIC) of the S&P 500 companies from 1993 to 2017
Median returns on invested capital (ROIC) of the S&P 500 constituents excluding financials from 1993 to 2017.
Goodwill is included.
Source: Morningstar Direct and own estimations.
Page 42 The results above are contrary to what McKinsey (2015, p. 104) found, which was a stable median ROIC from 1963 until the early 2000s at about 10% and an increase to 16% towards 2013. One might suspect that McKinsey’s sample would not only include large U.S. companies, such that smaller, fast growing companies and creative destruction could play a role in offsetting the lower returns of larger and maturing firms - but this is not the case. McKinsey utilized a comparable dataset of over 1.000 U.S.-based nonfinancial companies with revenues greater than
$1 billion to sort out small companies. So the different results should not be due to a size difference. Rather, the primary explanation is the included goodwill in our estimate of invested capital. The accumulated goodwill from M&A activity has significantly impacted the returns on invested capital negatively of the large U.S. firms since 1992 (McKinsey, 2015, p. 112).
Consequently, by including goodwill in ROIC, our valuations will be lower – especially for stocks that have carried out large acquisitions and accumulated goodwill on their balance sheets.
Summing up, future returns on invested capital should either move towards a common industry level, trend towards cost of capital (WACC) or remain close to a company-specific through-the-cycle level.
Valuation methods
To perform a quantitative valuation with the value driver formula, we have applied 4 different methods to estimate the inputs in the numerator. We test several variations of these to show that our results are robust across the different measures.
1. In the first and most basic value driver model, we estimate the equity value of firms by assuming last year’s NOPAT to grow in perpetuity while returns on new invested capital equal their sector’s cost of capital (WACC). We refer to this model as RONIC=WACC.
𝐸𝑞𝑢𝑖𝑡𝑦 𝑣𝑎𝑙𝑢𝑒𝑡 =𝑁𝑂𝑃𝐴𝑇𝑡−1 × (1 + 𝑔) × (1 − 𝑔
𝑅𝑂𝑁𝐼𝐶 = 𝑊𝐴𝐶𝐶)
𝑊𝐴𝐶𝐶 − 𝑔 − 𝑁𝐼𝐵𝐷𝑡
2. In a second, less conservative, model we estimate the equity value by assuming last year’s NOPAT to grow in perpetuity, and that the company can conduct new investments at the same return as last year’s ROIC. We refer to this model as LY ROIC.
𝐸𝑞𝑢𝑖𝑡𝑦 𝑣𝑎𝑙𝑢𝑒𝑡=
𝑁𝑂𝑃𝐴𝑇𝑡−1 × (1 + 𝑔) × (1 − 𝑔 𝑅𝑂𝐼𝐶𝑡−1)
𝑊𝐴𝐶𝐶 − 𝑔 − 𝑁𝐼𝐵𝐷𝑡
Page 43 3. In the third model we assume last year’s NOPAT to grow in perpetuity and that the firm’s return on new investments equal a median historical ROIC. We perform tests with 3-, 5-, and 10-year medians of ROIC. We refer to these models as 3Y median ROIC, 5Y median ROIC and 10Y median ROIC.
𝐸𝑞𝑢𝑖𝑡𝑦 𝑣𝑎𝑙𝑢𝑒𝑡 =𝑁𝑂𝑃𝐴𝑇𝑡−1 × (1 + 𝑔) × (1 − 𝑔
𝑀𝑒𝑑𝑖𝑎𝑛 𝑅𝑂𝐼C)
𝑊𝐴𝐶𝐶 − 𝑔 − 𝑁𝐼𝐵𝐷𝑡
4. The fourth approach is to estimate the steady-state NOPAT as a historical average while applying a historical median ROIC as a proxy for the steady-state return on new invested capital. We test the efficacy of 3-, 5-, and 10-year average NOPAT and median ROIC.
We refer to these as 3Y Average NOPAT, 5Y Average NOPAT, and 10Y Average NOPAT.
𝐸𝑞𝑢𝑖𝑡𝑦 𝑣𝑎𝑙𝑢𝑒𝑡= 1
𝑛∑ 𝑁𝑂𝑃𝐴𝑇𝑖 × (1 + 𝑔) × (1 − 𝑔
𝑀𝑒𝑑𝑖𝑎𝑛 𝑅𝑂𝐼𝐶)
𝑛𝑖=1
𝑊𝐴𝐶𝐶 − 𝑔 − 𝑁𝐼𝐵𝐷𝑡
A common feature of the value driver models described above is that the fundamental inputs are based on past financial performance and not subjective forecasts. The most recent year’s NOPAT is applied in the first three models due to being more relevant (Nissim, 2017, p. 16) and is relatively robust when adjusting for unusually large, small, and negative tax rates as described above.
In addition to being theoretically more valid and realistic to assume steady state RONIC equal to WACC in the first value driver model, it is also a conservative assumption for firms earning returns above their cost of capital while being an aggressive assumption for firms earning lower returns on capital. A single-period valuation is already quite conservative for fast growing companies with moats by assuming a low constant growth and not having an explicit forecast period or a prolonged period of sustained competitive advantages before steady state is reached (McKinsey, 2015, p. 256). Such qualitative valuation maneuvers would normally act as a runway for growth and increasing profitability before finally calculating terminal value. As a result, the RONIC=WACC model should not provide very attractive or precise valuations of fast-growing companies that have not yet matured but might only be suitable for stable and mature firms (McKinsey, 2015, p. 144). As we will illustrate later, relatively few firms are typically deemed undervalued by the first model compared to the others. Since we apply the same WACC to every company in a given sector, the assumption of RONIC to equal WACC will over time move the overall profitability (ROIC) of all firms within that sector to a common level. As these firms grow and make new investments with identical RONIC, this will gradually dilute the companies’
overall ROIC and make it converge towards the assumed sector WACC.
Page 44 In the second and third model we try to compensate for the conservative nature of the single-period valuation by assuming some less conservative inputs. We assume the return on new investments (RONIC) to be equal to a median of the firm’s 3-, 5-, or 10-year ROIC or simply the last year’s ROIC. This allows high-quality firms with competitive advantages and abnormal profitability to sustain these levels forever - subsequently increasing their terminal value. On the other hand, firms which have historically earned a ROIC below WACC, will not see their returns on capital rise to a value-neutral point (as our first model assumes). Instead, we assume these firms to stay the course and forever destroy shareholder value when they invest and grow (McKinsey, 2015, p. 22). Thereby, applying historical levels of profitability can prove a double-edged sword by overrating quality metrics and undervaluing junk stocks as defined by Pedersen, Asness & Frazzini (2013), although empirical studies indicate that profitability is sticky over time (McKinsey, 2015, p. 110 and Nissim, 2017, p. 17). A benefit of utilizing a historical median of ROIC is that these studies also point towards returns on capital exhibiting mean reversion over time. A longer-term median will also reflect the returns on capital through a whole economic cycle and include events of low returns such as The Great Recession and the Dot-com bubble.
In the fourth model we assume steady state NOPAT to equal a historical average instead of simply last year’s NOPAT. Assuming a historical average NOPAT will prove more conservative for companies that have experienced growth or improved their margins, as these companies’
most recent operating profit will be higher than their historical levels. On the other hand, it benefits firms with deteriorating businesses that have experienced declines in their operating profit. Thus, this model might favor distressed value stocks (Chen & Zhang, 1998, p. 532). This will also enable the model to evaluate firms that had a negative NOPAT last year. This method would prove more effective, if the nominal EBIT or NOPAT exhibits mean reversion in the long term just as the literature argues for margins and returns on capital. Combining an average of the past operating profit with the median of past returns on invested capital also presents a more consistent approach to estimating the inputs, as both NOPAT and ROIC is measured over the same time horizon. Combining a 10Y median ROIC with last year’s NOPAT, as we do in the 10Y median ROIC might not be a realistic assumption if a firm has grown its operating profit by investing heavily, as this would have increased NOPAT at the expense of ROIC. Growth requires investments, and assuming a high recent operating profit to compound at the high historical returns on capital would be too optimistic in the previous example. In this case, the fourth model would provide a more conservative assumption to the level of NOPAT in steady state.
Backtesting the value driver models with different assumptions for growth, WACC and fundamental inputs will illustrate whether the investment performance is solid across these variations.
Page 45